A page in a Google AI Overview does not have to appear on the first page of familiar blue links for the same query. Ahrefs measured that overlap in a March 2, 2026 analysis of about four million citation URLs across 863,000 keyword search results.
The headline rounds 37.9% from a comparison that includes search result features. With only ordinary blue links, the first-ten share was 37.1%. The distinction matters if you want to reproduce the comparison or decide what to track for your own pages.
- 0137.9% across result blocksAhrefs matched AI Overview citation URLs to the first ten result blocks for the same queries, including features outside the blue links.
- 0237.1% across blue linksA second analysis restricted the comparison to standard blue links. About 36.7% of citations were outside its top 100.
- 03Overlap is a share of citationsA cited page’s rank and a ranked page’s chance of being cited are different measurements.
- 04Record rankings and citations togetherCollect each observation against the same query, location, device and time so the two views can be compared.
01 — MethodStart with what Ahrefs matched
Louise Linehan’s Ahrefs study, Update: 38% of AI Overview Citations Pull From The Top 10, compared individual URLs cited in an AI Overview with the search results for the original query. Its data came from Ahrefs Brand Radar. A match meant the same URL appeared in both places for that query. That definition matters for companies with more than one useful page about a topic.
For example, an Overview might cite a company’s guide to sizing a product while the ordinary results contain the company’s product page. The company is visible in both views, and its individual URLs do not overlap. Keep both domain and URL measures if you track the question yourself. You may want the URL comparison for diagnosis and a domain view to track whether customers encounter a source from your business.
02 — ResultsRead the two distributions
In the all-block analysis, 37.9% of cited URLs were in the first ten result blocks, 31.2% were in blocks 11 to 100, and 31.0% were beyond the first 100 blocks. Ahrefs counted items such as featured snippets, People Also Ask, video packs and ads as blocks alongside organic listings. The published figures round independently.
In the blue-link-only analysis, 37.1% were in the first ten, 26.2% were in positions 11 to 100 and 36.7% were outside the first 100. Ahrefs also reported that YouTube made up 18.2% of cited URLs outside the first 100 in its reported non-ranking group, and 5.6% of all cited URLs in its dataset. Keep the denominator with each figure when discussing where citations appeared.
03 — InterpretationCheck the denominator
Suppose a search triggers an Overview with ten citations, and four of the cited URLs also rank in the top ten blue links. In that example the overlap is four out of ten cited URLs. A different study might start from the page in position one for every query and measure how often that page gets cited. Ahrefs’ aggregate number does not answer that question.
Ahrefs compared this update with its July 2025 study, which reported a higher share. The author also said Ahrefs improved its citation parsing for the later study. Treat the figures as snapshots with a method change between them. Google’s documentation for AI features describes query fan-out across related searches. That could help explain why a URL appears as a citation despite being absent from the direct results. The study did not isolate the cause of individual selections.
04 — Content reviewInspect the source page
Choose a question that matters to a reader and compare the cited pages with the ones you already publish. Open each source and check the claim it can support. A product page may explain a compatibility rule, a manual may show exact steps, and a support page may document a return exception. They serve different reader needs even where all three mention the same product.
Use this comparison to decide which facts or examples your page is missing. If a question is about fit, put the actual dimensions and supported variants in clear text. If the answer changes by version or date, identify which release the explanation covers. A technical SEO audit can help confirm that your strongest pages can be crawled and indexed.
05 — SamplingGroup questions by customer need
Ahrefs’ published report does not break the result out by industry, and its aggregate share cannot establish a healthcare, finance or ecommerce average. Build a relevant panel if you need to make decisions for one industry. For a shop, include common product research questions and comparison questions, then add queries about shipping or returns if customers use Search to resolve them. Preserve the entries even when they stop triggering an Overview.
Track each group separately and report the number of observations. A change from one cited query to two is a large relative change in a ten-query group. The count helps the team judge whether the change warrants investigation.
Observation
Record the cited URLs and ordinary links for the same query, time, device and locale.
Overlap
Match individual URLs consistently. Log domain overlap separately if it matters to your decision.
Performance
Use impressions and clicks to check actual web search traffic. A citation is an appearance.
06 — ImplementationChoose changes you can verify
Google’s guidance on generative Search requires no special markup for AI Overviews or AI Mode. A separate Ahrefs study, We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved., tracked URLs adding JSON-LD and compared them with a matched group. It reported little overall citation movement across Google AI Overviews, AI Mode and ChatGPT.
Keep accurate structured data for the features your page actually qualifies for. Refer to our JSON-LD implementation guide for examples. If your measurement panel shows that users need more detail about one product, verify the product facts and publish the detail on the relevant page. Record the template and release date so you can see which article or product changed.
07 — MeasurementCapture the same search context
For each observation, save the date, query, locale, device, presence of an Overview, cited URLs and blue-link results. Define how you handle URLs with tracking parameters or fragments so that the same document is counted consistently. Keep a copy of the raw observation before grouping by host, and mark the cases where the Overview was absent to preserve the denominator. For example, results collected on a desktop in Toronto and on a phone in Sydney belong in separate groups. Mixing them may make a locale change look like a gain or loss in citation coverage.
Google describes AI feature data as part of the Web search type in the Search Console Performance report. Use that report for web-search impressions and clicks. Use your search sample to count observed citations. A click or session needs its own measurement, so avoid treating the appearance of a citation as evidence of a visit. Our AI referral traffic analysis looks at referral traffic as its own metric.
Work through a small sample
Suppose you track twenty buyer questions for a backpack catalogue. Twelve trigger an Overview on a particular collection day. Four of those Overviews cite at least one URL on your site. Record four out of twelve as your domain’s citation rate among observed Overviews. If your goal is visibility across the whole question set, four out of twenty answers a different question. Report both counts and keep the wording consistent in the next comparison. Keep a record of changes to the query list.
Now open each of the four citations. If three point to the same sizing guide and one points to a product page, you have four cited query observations and two distinct cited URLs. That distinction matters when deciding what to maintain. A change to the sizing guide could affect three observations. A report that aggregated only the two URLs would hide that concentration.
Repeat the same measurements after a release and record which queries stopped triggering an Overview. If one month has more citations simply because more queries showed an Overview, that change affects how you read the result. Check the cited links and your Search Console records separately before asking whether the work brought qualified buyers to the site.
Larger archives can reuse the sample when evaluating template-wide releases. Keep a list of the affected templates, and track the first confirmed crawl after the release so that a recrawl delay is visible in the measurement. Our search services cover technical and editorial reviews for teams that need to connect these data.
08 — ComparisonKeep a baseline for the next read
Preserve the query panel and counting rules before you change a site. If you adopt a different tool, run both over a short overlap period and compare their outputs. A new parser may capture citations that the previous tool missed. Name the change alongside the metric, as you would when you update a rank-tracking tool or analytics definition.
Decide which action each measurement can inform. If citations routinely point to a source with facts absent from your indexed page, check whether those facts belong on the page and can be verified. If high-value pages disappear from the ordinary results, inspect their search performance and technical health too. The search visibility strategy guide covers how to work those findings into an editorial review.
Compare URLs on a consistent query set
Ahrefs’ result describes the overlap in its own Google sample. Track comparable observations over time to learn which cited pages matter for the queries your readers use.
Corrections
Correction, : We corrected the study attribution to Ahrefs and restored its reported sample and result breakdowns. The earlier version attributed rankings by query and industry, a cross-platform sample, an engagement effect and a 3.2-fold markup advantage to a study that did not measure those outcomes. We removed those claims and replaced the implementation advice with a workflow readers can use to track their own search visibility.